Elijah Newren <new...@gmail.com> writes: >> I'd expect that a reader of the commit who cares enough to bother to >> wonder by looking at the patch and seeing that 2 became 3 would know >> why already. And a reader of the resulting file would not know that >> the 3 used to be 2, and won't be helped by "we used to count to 2, >> now we have 'out' also counted" that much, especially in the commit >> log message. What would help the latter would be to name which >> three paths we expect to see in the comment (or test against the >> exact list of paths, instead of using test_line_count). >> >>> An alternative to consider would be to add a .gitignore file in the >>> initial commit to ignore 'out', then the number of untracked files >>> don't have to be adjusted. >> >> I think that is a preferred solution that we've used in ls-files and >> status tests successfully. > > ...except that if we add a .gitignore to each initial commit (we use > test_create_repo for nearly every test to keep them separable meaning > we'd have to do this many times), then four lines above we have to > adjust the number of expected tracked files. And, for it to work, > we'd have to add an --exclude-standard flag to ls-files -o.
Yeah, unless the original planned to use the .gitignore mechanism, converting it to use it now will become noisy.
